baalajimaestro/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

fix indexing token_logprobs after sorting

Browse source
This commit is contained in:
wu-qing-157 2023-07-07 10:18:49 +00:00
parent ca11673061
commit 9e61661518
1 changed files with 3 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
llama_cpp/llama.py
View file

@ -958,7 +958,7 @@ class Llama:
)
],
"text_offset": [text_offset],
"token_logprobs": [sorted_logprobs[int(token)][0]],
"token_logprobs": [current_logprobs[int(token)]],
"top_logprobs": [top_logprob],
}
returned_tokens += 1
@ -1033,7 +1033,7 @@ class Llama:
self.detokenize([token]).decode("utf-8", errors="ignore")
],
"text_offset": [text_offset],
"token_logprobs": [sorted_logprobs[int(token)][0]],
"token_logprobs": [current_logprobs[int(token)]],
"top_logprobs": [top_logprob],
}
@ -1131,7 +1131,7 @@ class Llama:
zip(logprobs_token, range(len(logprobs_token))), reverse=True
)
)
token_logprobs.append(sorted_logprobs[int(token)][0])
token_logprobs.append(logprobs_token[int(token)])
top_logprob: Optional[Dict[str, float]] = {
self.detokenize([i]).decode("utf-8", errors="ignore"): logprob
for logprob, i in sorted_logprobs[:logprobs]